V1.1.0

.gitattributes

@@ -0,0 +1,2 @@
+# SCM syntax highlighting & preventing 3-way merges
+pixi.lock merge=binary linguist-language=YAML linguist-generated=true

.gitignore

@@ -20,4 +20,6 @@ __pycache__
 env
 data
 project
-!project/*.sh
+!project/*.sh# pixi environments
+.pixi/*
+!.pixi/config.toml

pyproject.toml

@@ -1,31 +1,24 @@
-[tool.poetry]
+[project]
+dependencies = ["scikit-learn>=1.7.2,<2", "typer>=0.14.0", "pandas>=1.5.3", "opencv-python-headless>=4.11.0.86", "pyheif>=0.8.0", "pillow>=11.3.0"]
 name = "smartsensor"
+requires-python = ">=3.9"
 version = "1.1.0"
-description = "The sensor utilizes a smartphone camera to analyze and characterize chemical compounds."
-readme = "README.md"
-authors = ["Thanh-Giang (River) Tan Nguyen <nttg8100@gmail.com>", "Dung Cao <dung.cao@ttu.edu.vn>"]
-license = "MIT"
-homepage = "https://github.com/riverxdata/smartsensor"
-repository = "https://github.com/riverxdata/smartsensor"
-keywords = ["machine learning", "sensor", "smartphone"]
 
-classifiers = [
-    "Programming Language :: Python :: 3",
-    "License :: OSI Approved :: MIT License",
-    "Operating System :: OS Independent"
-]
-[tool.poetry.dependencies]
-python = ">=3.9"
-typer = ">=0.14.0"
-scikit-learn = ">=1.6.1"
-pandas = ">=1.5.3"
-opencv-python-headless = ">=4.11.0.86"
-streamlit = ">= 1.45.1"
-pyheif = ">=0.8.0"
-pillow = ">=11.3.0"
-[tool.poetry.scripts]
+[build-system]
+build-backend = "hatchling.build"
+requires = ["hatchling"]
+
+[tool.pixi.workspace]
+channels = ["conda-forge", "bioconda"]
+platforms = ["linux-64"]
+
+[project.scripts]
 smartsensor = "smartsensor.main:app"
 
-[build-system]
-requires = ["poetry-core>=1.0.0"]
-build-backend = "poetry.core.masonry.api"
+[tool.pixi.pypi-dependencies]
+smartsensor = { path = ".", editable = true }
+
+[tool.pixi.tasks]
+
+[tool.pixi.dependencies]
+libheif = ">=1.19.7,<2"

smartsensor/e2e.py

@@ -3,6 +3,7 @@
 from smartsensor.model.train import fit
 from smartsensor.model.metric import evaluate_metrics, aggregate_replication_metrics
 import pandas as pd
+import numpy as np
 import os
 import tempfile
 import warnings
@@ -23,7 +24,10 @@ def end2end_pipeline(
     prefix: str,
     test_size: float = 0.2,
     replication: int = 1000,
+    seed: int = 1,
 ):
+    # set seed for reproducibility
+    np.random.seed(seed)
 
     metrics = []
     features = features.split(",")

smartsensor/main.py

@@ -8,9 +8,7 @@
 from smartsensor.process.any2jpg import heic2jpg as convert_heic_to_jpg
 
 
-app = typer.Typer(
-    help="Smart Optical Sensor: Using smartphone camera as sensor", add_completion=False
-)
+app = typer.Typer(help="Smart Optical Sensor: Using smartphone camera as sensor", add_completion=False)
 
 
 @app.command()
@@ -38,12 +36,11 @@ def model(
         "--cv",
         help="Fold for cross validation",
     ),
-    test_size: float = typer.Option(
-        0.2, "--test-size", help="Test data size for splitting to train the model"
-    ),
+    test_size: float = typer.Option(0.2, "--test-size", help="Test data size for splitting to train the model"),
     degree: int = typer.Option([1], "--degree", help="Degree of polynomial regression"),
     replication: int = typer.Option(100, "--replication", help="Number of replication"),
     out: str = typer.Option(".", help="Folder to save model"),
+    seed: int = typer.Option(1, "--seed", help="Random seed for reproducibility"),
 ):
     """
     Run the end-to-end model for Smart Optical Sensor.
@@ -58,6 +55,7 @@ def model(
     logger.info(f"Test size: {test_size}")
     logger.info(f"Degree: {degree}")
     logger.info(f"Output folder: {out}")
+    logger.info(f"Random seed: {seed}")
     os.makedirs(out, exist_ok=True)
     end2end_pipeline(
         data=data,
@@ -71,6 +69,7 @@ def model(
         prefix=prefix,
         test_size=test_size,
         replication=replication,
+        seed=seed,
     )
 
 
@@ -84,6 +83,7 @@ def process(
         help="Folder to save processed images",
     ),
     process_dir: str = typer.Option(
+        None,
         "--process-dir",
         help="Path to the processed to get config for base background color",
     ),

smartsensor/model/metric.py

@@ -11,7 +11,7 @@ def evaluate_metrics(
     data: DataFrame,
     features: List,
     degree: int,
-) -> (DataFrame, DataFrame):
+) -> tuple[DataFrame, DataFrame]:
     """Simple evaluation matrics for measure the errors
 
     Args:
@@ -41,12 +41,8 @@ def evaluate_metrics(
     detail = detail.T
     detail.columns = ["image", "expected_concentration", "predicted_concentration"]
     detail = detail.sort_values("expected_concentration")
-    detail["absolute_error"] = abs(
-        detail["predicted_concentration"] - detail["expected_concentration"]
-    )
-    detail["error"] = (
-        detail["predicted_concentration"] - detail["expected_concentration"]
-    )
+    detail["absolute_error"] = abs(detail["predicted_concentration"] - detail["expected_concentration"])
+    detail["error"] = detail["predicted_concentration"] - detail["expected_concentration"]
 
     return (metric, detail)
 

smartsensor/model/split_data.py

@@ -9,22 +9,21 @@ def split_data(
     prefix: str,
     test_size: float = 0.2,
 ):
-    """It will combine the features and metadata, split to prepare for training and testing
+    """It will combine the features and metadata, split to prepare for training and testing according to concentration.
 
     Args:
         meta_data(pd.DataFrame): The dataframe with data and metadata
         outdir (str): The output directory for train and test csv file
         prefix (str): The prefix to specify csv file
-        random_state (int, optional): To reproduce split. Defaults to 1.
         test_size (float, optional): The test size if train, test batch is not specified. Defaults to 0.2.
     """
-    # train test split data
+    # Train test split data
     # Define train and test data holders
     train_data = pd.DataFrame()
     test_data = pd.DataFrame()
     # split dataset by batch equally
     if test_size != 0:
-        for dataset_name, group_df in meta_data.groupby("concentration"):
+        for _, group_df in meta_data.groupby("concentration"):
             train_subset, test_subset = train_test_split(group_df, test_size=test_size)
             train_data = pd.concat([train_data, train_subset])
             test_data = pd.concat([test_data, test_subset])

smartsensor/process/normalize.py

@@ -54,19 +54,15 @@ def normalize(
     with open(os.path.join(outdir, "config.json"), "w") as f:
         json.dump({"lum": lum}, f)
 
-    with open(failed_ratio_file, "w") as f_ratio, open(
-        failed_delta_file, "w"
-    ) as f_delta:
+    with open(failed_ratio_file, "w") as f_ratio, open(failed_delta_file, "w") as f_delta:
         f_ratio.write("image\n")
         f_delta.write("image\n")
 
         for image_location in glob.glob(os.path.join(raw_roi, "*.jpg")):
             file_name = os.path.basename(image_location)
             background_image_path = os.path.join(background, file_name)
             if not os.path.exists(background_image_path):
-                raise ValueError(
-                    f"Background image {background_image_path} does not exist."
-                )
+                raise ValueError(f"Background image {background_image_path} does not exist.")
             raw_roi_img = cv2.imread(image_location)
             background_img = cv2.imread(background_image_path)
             logger.info(
@@ -194,7 +190,7 @@ def normalize_delta(
     deltaG = lum[1] - G
     deltaR = lum[2] - R
 
-    tmp = roi_image
+    tmp = roi_image.astype(np.float64)
     tmp[:, :, 0] += deltaB
     tmp[:, :, 1] += deltaG
     tmp[:, :, 2] += deltaR
@@ -219,11 +215,7 @@ def normalize_delta(
 
     logger.info(f"Normalize delta: {deltaB, deltaG, deltaR}")
 
-    if (
-        abs(deltaB) > THRESHOLD_DELTA
-        or abs(deltaG) > THRESHOLD_DELTA
-        or abs(deltaR) > THRESHOLD_DELTA
-    ):
+    if abs(deltaB) > THRESHOLD_DELTA or abs(deltaG) > THRESHOLD_DELTA or abs(deltaR) > THRESHOLD_DELTA:
         return True
 
     return False